Cartesian contact force estimation for robotic manipulators using Kalman filters and the generalized momentum

Estimating contact forces and torques in Cartesian space enables force-controlled robotic applications as well as collision detection without costly additional sensing. A new approach towards online estimation of contact forces and torques at the tool center point from motor torques as well as joint angles and speeds is presented, which is based on analyzing the generalized momentum of the manipulator. Existing generalized momentum based methods are extended by designing a Kalman filter to estimate the generalized momentum as well as contact forces and torques simultaneously. This approach introduces additional degrees of freedom in the design of the estimator that are exploited to increase robustness with respect to disturbances such as uncertainty in joint friction. The method is verified by simulation results obtained from a dynamic model of an ABB YuMi, a dual-arm collaborative robot with 7DOF each arm.